Global Speed Planning Algorithm for Intelligent Connected Buses Based on DDPG with Adaptability to Vehicle Mass

Intelligent connected bus (ICB) can obtain traffic light phase sequence information and plan the global speed in advance, so as to achieve fast, energy-saving, and comfortable traffic. However, traditional methods have limitations in vehicle mass adaptation, optimization ability, real-time calculation speed and other aspects. Therefore, a global speed planning algorithm of adaptive vehicle mass based on deep deterministic policy gradient (DDPG) is proposed in this paper, which covers the typical scenes of intersections with signal lights and entering and leaving stations. First, an electric bus powertrain and vehicle dynamics model based on a simplified operation mechanism are established, which provides an object model for vehicle speed planning algorithm. Then, the change in bus speed is modeled as the Markov decision process (MDP) and the idea of expanding the state of bus mass is put forward, combined with the estimation of the vehicle mass on the line, so as to realize the active adaptation of the vehicle speed to the passenger capacity. A reward function is designed that takes into account travel energy consumption, traffic time, ride comfort, and other factors. Finally, the algorithm is tested and verified according to the topology of intelligent connected open road and traffic light phase sequence in Tianjin Haihe Education Park (THEP). The results show that, compared with the rule-based method, the total reward of the proposed algorithm increases by 10.1% on average, and can reach about 97.3% of dynamic planning (DP). After the off-line training, the calculation time of the proposed algorithm is reduced by 86.38% on average compared with DP, and the calculation speed is faster. And it has good generalization ability, can adapt to changes in scene and vehicle weight, and has the potential for real vehicle deployment.